> For example, when the GPU is fully idle, nvidia-smi tells me that it’s only pulling 88W of power.

I haven't used a non-laptop GPU in some time, but that is a crazy amount of "idle" power consumption. Is this normal for cards like this?

Server cards are not optimized for idle power usage. They’re expected to be fully utilized.

For server gear it’s more common to have less dynamic power and voltage switching because it produces more predictable performance and latency.

For GeForce cards you can get similar behavior by setting “Prefer maximum performance” which disables some of the low power states.

CPU sleep states are normally disabled on servers because CPU demand can increase faster than the CPU speed can increase. It is much better to have a $20,000 CPU running at max MHz at all times.

If my gpu is sitting idle, and I mean idle with nothing loaded into its memory, it's sitting at about 18W. If I load in model that uses nearly all of the memory but that model is idle, it's at 36W. If that model is actively thinking, it's like 118W. I think this is likely due to the GPU being aware that there is real data loaded into memory and turning up the DRAM refresh rate whereas when nothing is loaded, the dynamic power is as low as possible.

Yes, I have some of these cards and AFAICT the HBM2e chips just always run at full speed. I have different variants of the pcie cards and while I can get the gpu itself into a lower power state the memory just runs full tilt. Though I see 40w on my “normal” cards and 60w on the Frankenstein card that thinks it’s an sxm4.

IIRC this was one of the issues with 2/2e, some combination of the various available memory controllers not agreeing on a standard to manage timings and power states. I haven't played around with my Radeon VII in a long while now.

That aside idle power consumption is a driver-to-driver affair from both amd and novideo, sometimes I'm only pulling 15-30W when nothing is happening and other times it decides it needs 110w for a static 500hz screen

I suspect the act of running nvidia-smi itself prevents the GPU from being put into a low-power state.

From memory this is true and nvml (Nvidia management library) is the way to get stats that doesn't cause the GPU to wake.

I believe they tried to inject some chemicals to slow the reaction, but the pump and/or valves failed and clogged.

I was thinking maybe have those chemicals sitting in a glass or temperature sensitive container inside the tank. So when there's too much pressure or heat, the container containing the neutralizing chemical is broken like a fuse and the chemical is automatically released.

Having the chemical in one location doesn't make it active all over, you need to disperse it. Like you need to shake glow-in-the-dark bracelets.

Well then... make a matrix of such fuse-containers? (say every 20cm or whatever) I guess manufacturing such a matrix would be pretty expensive though, you'd need to carefully automate its production I think. It would also definitely interfere with flow of fluid in the tank.

I was thinking multiple long skinny tubes with etchings that make them more likely to split lengthwise. Maybe with a spring loaded/powered agitator so when the tube breaks there's some mechanical flinging/mixing of the inner chemical.

But I'm not a chemical processes engineer, so I don't know how much mixing is needed. But the existing emergency plan was to inject the the chemical through a single valve, so it seems like the dispersion and mixing requirements in this case seems to be low.

Easy to imagine protective mechanisms in hindsight, now imagine you have to build a system for every tank of every chemical in the world

I am imagining some sort of regulations that enforce the creation and installation of appropriate safety systems upon all similar sites in the affected jurisdiction.

And a standard to prevent similar danger elsewhere in the world.

Or is that too fantastic? Perhaps I should stick with the distopian fiction I so love. Why haven't flying cars taken off yet?

That is active.

Something passive could be submerging the tank in a pool of water (also good for proving spill containment won't leak).

Typically you don't have enough surface area for that. The walls are thick enough that thermal conductivity into an ambient-temperature liquid alone is not going to be sufficient.

I think the passive version is the tank stays in a pool all the time.

Riiight. That is exactly what I was thinking.

It's really unclear what you were thinking

My first MP3 player used it as well. I remember how much this memory card style felt like a 5¼ floppy disk, as opposed to most other formats at the time which were hard shell.

> As a side effect, Terms-of-service acceptance records were also reset, prompting users to re-accept on their next visit to the dashboard.

Don't get me wrong- the rest of this mess falls pretty clearly on Google Cloud, but this one feels like something Railway did to themselves.

I wanted to find this interesting, but it has AI/LLM signs of writing all over it.

The dig in the middle - "you can skip the next part, but if you do skip, are you even a real reader? Not judging. Just saying" - ugh. Why would I bother reading every word if you likely didn't write every word?

I can assure you that every word was written (or dictated) by a human. At least you read up to that point in the article. But did you read beyond that.

This reads like LLM generated post. "actually" is in three of the headers, among other signs. https://tropes.fyi/

Same. Overuse of the word "genuinely" is another tell.

You've clearly never read my writing. I overuse "actually" all the time.

Maybe he actually read it.

This is an Internationalized domain name (https://en.wikipedia.org/wiki/Internationalized_domain_name) - the actual domain shows up in browsers as マリウス.com but HN unfortunately doesn't decode or show it that way.

I wrote up our experience of how to get Tiptap to play nice with a Go backend server, along with a lot of inline code snippets and a gist with the backend code. Happy to answer any other questions I may have missed in writing it up.

Even if you don't use the Tiptap editor, hopefully the idea of a sidecar container is helpful to someone trying to solve a similar problem of allowing Typescript/Node code to interact with Go.

Wow, did not know this site was still online. I apparently have a user ID in the 200s, was an early sign up to the site. https://www.ticalc.org/community/directory/

Good luck, and I hope it works out! Make sure you are ready to ride the roller coaster of highs and lows, as there are going to be many. Remember that your time and experience are the most valuable things you have - make sure you're in control of both.